SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2008.08692 · 2020

Enhancing Graph Neural Network-based Fraud Detectors against Camouflaged Fraudsters

arXiv · PDF · Open in the Atlas

Code that ran

12 of 17 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 2 failures of our instrument; 5 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2008.08692.json · badge: /api/ran/2008.08692.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
YingtongDou/CARE-GNN canonical (classifier label)
classifier basis: census
8 ran · 2 no recorded run
safe-graph/dgfraud-tf2 community (classifier label)
classifier basis not recorded on the link
4 ran · 3 no recorded run
FunctionStatusWhere it lives
build_batch Ran
ran on a synthesized input
safe-graph/dgfraud-tf2/algorithms/GraphSage/GraphSage_main.py
code served (permissive licence) · get_code("afac081736ef9604")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned MiniBatch. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import collections

np.random.seed(0)

# Create a small graph with bidirectional neighbors
# Nodes: 0, 1, 2, 3, 4
neigh_dict = {
    0: np.array([1, 2]),
    1: np.array([0, 2, 3]),
    2: np.array([0, 1, 4]),
    3: np.array([1, 4]),
    4: np.array([2, 3]),
}

# Target nodes to build batch for
nodes = np.array([3, 4], dtype=np.int64)

# Sample sizes for each layer (2 layers)
sample_sizes = [2, 2]

ARGS = {
    'nodes': nodes,
    'neigh_dict': neigh_dict,
    'sample_sizes': sample_sizes,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MiniBatch — a MiniBatch output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_diffusion_matrix Ran
ran on a synthesized input
safe-graph/dgfraud-tf2/algorithms/GraphSage/GraphSage_main.py
code served (permissive licence) · get_code("c6c6448a408df9a3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

np.random.seed(0)

# Construct a small neighbor dictionary
# Keys are node IDs, values are lists of neighboring node IDs
neigh_dict = {
    0: np.array([1, 2, 3]),
    1: np.array([0, 2]),
    2: np.array([0, 1, 3, 4]),
    3: np.array([0, 2]),
    4: np.array([2]),
    5: np.array([1, 3, 4]),
}

# Destination nodes to compute diffusion matrix for
dst_nodes = np.array([0, 2, 5])

# Sample size for neighbors
sample_size = 2

# Maximum node ID in the graph
max_node_id = 5

ARGS = {
    "dst_nodes": dst_nodes,
    "neigh_dict": neigh_dict,
    "sample_size": sample_size,
    "max_node_id": max_node_id,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
date_diff Ran
ran on a synthesized input
YingtongDou/CARE-GNN/amazon_preprocess.py
code served (permissive licence) · get_code("2f5a6dba21e8cc75")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import time

# Get two unix timestamps representing different dates
# Using time.time() to get current timestamp, then subtracting seconds for a past timestamp
current_timestamp = time.time()
past_timestamp = current_timestamp - (5 * 24 * 60 * 60)  # 5 days ago

ARGS = {
    "t1": current_timestamp,
    "t2": past_timestamp,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
dot Ran
ran on a synthesized input
safe-graph/dgfraud-tf2/layers/layers.py
code served (permissive licence) · get_code("4c699fbcd8522a90")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned EagerTensor, ndim 2, dtype family float, shape (2, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf

# Case 1: dense matmul (sparse=False)
x_dense = tf.constant([[1.0, 2.0, 3.0, 4.0],
                       [5.0, 6.0, 7.0, 8.0]], dtype=tf.float32)  # shape (2, 4)
y_dense = tf.constant([[1.0, 2.0, 3.0],
                       [4.0, 5.0, 6.0],
                       [7.0, 8.0, 9.0],
                       [10.0, 11.0, 12.0]], dtype=tf.float32)  # shape (4, 3)

# Case 2: sparse matmul (sparse=True)
indices = [[0, 0], [0, 2], [1, 1], [1, 3]]
values = [1.0, 2.0, 3.0, 4.0]
shape = [2, 4]
x_sparse = tf.SparseTensor(indices=indices, values=values, dense_shape=shape)

ARGS = {
    "x": x_dense,
    "y": y_dense,
    "sparse": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
filter_neighs_ada_threshold Ran
ran on a synthesized input
YingtongDou/CARE-GNN/layers.py
code served (permissive licence) · get_code("a9d2dcd649e1c69d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Construct arguments for filter_neighs_ada_threshold
# The function processes batch nodes with their scores and neighbors

batch_size = 2
num_neighbors_per_node = 3

# center_scores: tensor of shape [batch_size, 1] containing label-aware scores for batch nodes
center_scores = torch.randn(batch_size, 1)

# neigh_scores: list of tensors, each of shape [num_neighbors, 1] for each batch node
neigh_scores = [
    torch.randn(num_neighbors_per_node, 1),  # scores for neighbors of batch node 0
    torch.randn(num_neighbors_per_node, 1),  # scores for neighbors of batch node 1
]

# neighs_list: list of lists containing neighbor node ids for each batch node
neighs_list = [
    [10, 11, 12],  # neighbor ids for batch node 0
    [20, 21, 22],  # neighbor ids for batch node 1
]

# sample_list: list of ints specifying how many neighbors to keep for each batch node
sample_list = [2, 2]

ARGS = {
    "center_scores": center_scores,
    "neigh_scores": neigh_scores,
    "neighs_list": neighs_list,
    "sample_list": sample_list,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
getdict Ran
ran on a synthesized input
YingtongDou/CARE-GNN/amazon_preprocess.py
code served (permissive licence) · get_code("9671520f763cedba")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import gzip
import tempfile
import os

# Create a temporary gzip file with test data
# The function expects a gzip file where each line is evaluable to a dict
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_data.gz")

with gzip.open(temp_file, 'wb') as f:
    # Write some test dictionaries, one per line
    # The first value in each dict becomes a key in the reviews dict
    f.write(b"{'product_id': 'A001', 'rating': 5, 'text': 'Great!'}\n")
    f.write(b"{'product_id': 'A001', 'rating': 4, 'text': 'Good'}\n")
    f.write(b"{'product_id': 'B002', 'rating': 3, 'text': 'OK'}\n")
    f.write(b"{'product_id': 'A001', 'rating': 5, 'text': 'Excellent'}\n")

ARGS = {
    "path": temp_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mean_inter_agg Ran
ran on a synthesized input
YingtongDou/CARE-GNN/layers.py
code served (permissive licence) · get_code("c80c8faa06ecf716")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn.functional as F

torch.manual_seed(0)

# Determine dimensions and sizes
num_relations = 3
n = 2  # number of nodes in batch
embed_dim = 4
feat_dim = 5

# Create self_feats: shape (n, feat_dim)
self_feats = torch.randn(n, feat_dim)

# Create neigh_feats: shape (num_relations * n, feat_dim)
# This is because the function indexes it as neigh_h[r * n:(r + 1) * n, :]
neigh_feats = torch.randn(num_relations * n, feat_dim)

# Create weight: shape (feat_dim, embed_dim)
# This is used in torch.mm(self_feats, weight) and torch.mm(neigh_feats, weight)
weight = torch.randn(feat_dim, embed_dim)

# Set cuda to False (CPU only)
cuda = False

ARGS = {
    "num_relations": num_relations,
    "self_feats": self_feats,
    "neigh_feats": neigh_feats,
    "embed_dim": embed_dim,
    "weight": weight,
    "n": n,
    "cuda": cuda,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize Ran
ran; our contract draft was wrong, not the code
YingtongDou/CARE-GNN/utils.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from blacksingular/wsdm_gdn, is pointer only (licence: NONE) · get_code("5d14a28a520ecbee")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from blacksingular/wsdm_gdn, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

normalize Ran
ran; our contract draft was wrong, not the code
YingtongDou/CARE-GNN/simi_comp.py
code served (permissive licence) · get_code("393c2aa36f176d58")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned csr_matrix. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import scipy.sparse as sp

# Create a small sparse matrix for testing
# The function expects a sparse matrix that can be row-normalized
# Using a 4x3 matrix to have distinct dimensions
data = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
row = np.array([0, 0, 1, 2, 3])
col = np.array([0, 1, 1, 2, 2])
sparse_matrix = sp.csr_matrix((data, (row, col)), shape=(4, 3))

ARGS = {
    "mx": sparse_matrix
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type csr_matrix — a csr_matrix output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a csr_matrix output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
pos_neg_split Ran
ran on a synthesized input
YingtongDou/CARE-GNN/utils.py
code served (permissive licence) · get_code("4f136d9fcf969069")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 2, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import copy as cp

torch.manual_seed(0)

# The function takes:
# - nodes: a list of nodes (can be any objects/values)
# - labels: a list of node labels (should be integers, checking for label == 1)

# Create simple nodes - can be integers, strings, or any hashable objects
nodes = [10, 20, 30, 40]

# Create labels - mix of 0s and 1s to test positive/negative split
labels = [0, 1, 0, 1]

ARGS = {
    "nodes": nodes,
    "labels": labels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
scaled_dot_product_attention Ran
ran on a synthesized input
safe-graph/dgfraud-tf2/layers/layers.py
code served (permissive licence) · get_code("00245cbc1ab1f183")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)

# Create small test tensors
# q: query tensor, shape (batch=2, seq_len=3, embed_dim=4)
q = tf.constant(np.random.randn(2, 3, 4).astype(np.float32))

# k: key tensor, shape (batch=2, seq_len=3, embed_dim=4)
# Must have same embedding dimension as q for matmul compatibility
k = tf.constant(np.random.randn(2, 3, 4).astype(np.float32))

# v: value tensor, shape (batch=2, seq_len=3, embed_dim=5)
# Can have different embedding dimension than q and k
v = tf.constant(np.random.randn(2, 3, 5).astype(np.float32))

ARGS = {
    "q": q,
    "k": k,
    "v": v,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
star_judge Ran
ran on a synthesized input
YingtongDou/CARE-GNN/amazon_preprocess.py
code served (permissive licence) · get_code("7785fece996c91c2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function iterates through reviews1 and reviews2, indexing each review with [0]
# So reviews1 and reviews2 should be lists of sequences (tuples, lists, or tensors)
# where the first element can be compared with ==

reviews1 = [
    (1, "great product"),
    (2, "okay quality"),
    (3, "not satisfied")
]

reviews2 = [
    (2, "average experience"),
    (4, "excellent service"),
    (5, "would buy again")
]

ARGS = {
    "reviews1": reviews1,
    "reviews2": reviews2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
RLModule Not yet run
unverified (harvested, no recorded run)
YingtongDou/CARE-GNN/layers.py
code served (permissive licence) · get_code("ff287257b0b7f3a6")
How it ran

no invocation recorded.

build_batch Not yet run
unverified (harvested, no recorded run)
safe-graph/dgfraud-tf2/algorithms/GraphConsis/GraphConsis_main.py
code served (permissive licence) · get_code("d3e0193f69dfa61c")
How it ran

no invocation recorded.

compute_diffusion_matrix Not yet run
unverified (harvested, no recorded run)
safe-graph/dgfraud-tf2/algorithms/GraphConsis/GraphConsis_main.py
code served (permissive licence) · get_code("848f8034ca5d8ab4")
How it ran

no invocation recorded.

load_data Not yet run
unverified (harvested, no recorded run)
YingtongDou/CARE-GNN/utils.py
code served (permissive licence) · get_code("2ab0f159b61e9e5b")
How it ran

no invocation recorded.

sparse_dropout Not yet run
unverified (harvested, no recorded run)
safe-graph/dgfraud-tf2/layers/layers.py
code served (permissive licence) · get_code("fcf4f04027e8dde1")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 15 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
normalize 1 other paper
1 later
2401.14155 (2024-01)
Alleviating Structural Distribution Shift in Graph Anomaly Detection

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Graph Neural Networks (GNNs) have been widely applied to fraud detection problems in recent years, revealing the suspiciousness of nodes by aggregating their neighborhood information via different relations. However, few prior works have noticed the camouflage behavior of fraudsters, which could hamper the performance of GNN-based fraud detectors during the aggregation process. In this paper, we introduce two types of camouflages based on recent empirical studies, i.e., the feature camouflage and the relation camouflage. Existing GNNs have not addressed these two camouflages, which results in their poor performance in fraud detection problems. Alternatively, we propose a new model named CAmouflage-REsistant GNN (CARE-GNN), to enhance the GNN aggregation process with three unique modules against camouflages. Concretely, we first devise a label-aware similarity measure to find informative neighboring nodes. Then, we leverage reinforcement learning (RL) to find the optimal amounts of neighbors to be selected. Finally, the selected neighbors across different relations are aggregated together. Comprehensive experiments on two real-world fraud datasets demonstrate the effectiveness of the RL algorithm. The proposed CARE-GNN also outperforms state-of-the-art GNNs and GNN-based fraud detectors. We integrate all GNN-based fraud detectors as an opensource toolbox: https://github.com/safe-graph/DGFraud. The CARE-GNN code and datasets are available at https://github.com/YingtongDou/CARE-GNN.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2008.08692")
get_code_for_paper("2008.08692")
have("2008.08692")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2008.08692.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2008.08692.svg)](https://syntology.ai/paper/2008.08692)

Connect an agent — have() is free.